Revealed: How HS2 has become too expensive to stop
Key Points
- NAO: HS2 would have scored 0.3–0.4 benefit–cost ratio ("poor value") if 2026 costs were known in 2020
- Cost to complete London–Birmingham now £87.7bn–£102.7bn, roughly double the 2020 estimate
- Current ratio of 1.5–6.4 props up only because cancelling now costs £33bn–£58bn
- Full railway delayed up to 13 years, to between May 2040 and December 2043
- £46.8bn already spent; reset due to complete spring 2027
HS2 would have been rated “poor value for money” if its true cost had been clear when ministers gave the final go-ahead in 2020, the National Audit Office has revealed.
In a report published on Monday (29 June), the spending watchdog set out the benefit–cost ratio the Department for Transport (DfT) would have arrived at in 2020 had it known the railway would cost what it now expects: between 0.3 and 0.4.
Anything below 1 means a project returns less than it costs, and DfT’s own guidance files that range under “poor value for money”. On those numbers, the case for building HS2 at all would have collapsed.
That is not the number the government is working with today.
DfT now puts the benefit–cost ratio for completing the London-to-Birmingham line at between 1.5 and 6.4, which it grades “medium to very high value for money”. The awkward part is how it gets there.
The ratio has actually improved even as the railway has become dramatically more expensive, because the sums only count money still to be spent, not money already gone.
And the estimated cost of cancelling HS2 has more than quadrupled since 2023. HS2 Ltd now reckons returning the land and unwinding contracts would cost £33 billion to £58 billion, roughly what it would cost to finish the thing.
When scrapping a project costs about as much as completing it, completing it always looks like value.
DfT, following HM Treasury Green Book rules, deducts that avoided cancellation cost from the bill, which flatters the ratio.
DfT told the NAO the government’s decision to press on rested on advice that the ratio cleared 1.5, rather than on the full range. HS2 Ltd itself admits the cancellation figure is a high-level estimate carrying “significant uncertainty”, with no comparable project to benchmark it against.
The numbers behind the reset
- Cost to complete London–Birmingham: £87.7 billion to £102.7 billion, roughly double the 2020 estimate
- Cost increase since 2020, excluding inflation: £35.9 billion to £37.2 billion, over 100% at the low end
- Spent so far to end March 2026: £46.8 billion, including £2.6 billion sunk into the cancelled Phase 2
- 2026 benefit–cost ratio for completion: 1.5 to 6.4
- Benefit–cost ratio if 2026 costs had been known in 2020: 0.3 to 0.4
The May 2026 figures, announced by Transport Secretary Heidi Alexander, also confirmed the line will open far later than promised. Initial services between Old Oak Common in west London and Birmingham are now expected between May 2036 and October 2039.
Full services running through to Euston and north of Birmingham onto the West Coast Main Line slip to between May 2040 and December 2043, up to 13 years later than the 2020 plan.
To claw back some of that, DfT has decided HS2 trains will run slower than designed, at a maximum of 320 km/h rather than 360 km/h.
That avoids the cost and time of testing the railway at the higher speed and could save £1 billion to £2.5 billion, though it knocks around £1.3 billion off the long-term benefits from slower journeys.
The watchdog stops short of saying the railway should not be finished. Its sharper warning is about discipline from here.
DfT and HS2 Ltd are not due to complete the current reset, the second since 2020, until spring 2027, and the NAO said they should not put revised plans into action until they are confident the cost baseline, commercial deals and Euston plans actually hold.
HS2 Ltd has already spent £101 million on the reset alone, against a £153 million budget.
The benefit–cost arithmetic now leans heavily on one uncomfortable fact: HS2 has reached the point where it is too expensive to stop.
